One of the things that amazes me about R is the incredible eco-system of packages that are available to make data analysis easier, and sometimes to even just make things that little more aesthetically pleasing that what base R plots can provide.
Much of what is posted within this blog is leveraging the incredible skills of people like Hadley Wickham, Yihui Xie, and Bob Rudis to name just a few. One of the great things about having such a rich eco-system, is that if there’s ever anything you’re wanting to do someone may have already written a package to help you do it. That is the case for this post making use of the
Rfacebook (CRAN, Github) package.
In my day job I work with a range of charities and look at wide assortment of data to help them understand how their programs are performing. One thing that we’ve observed over the last two years, is that the core methods we’ve used to recruit donors historically (I.e. Direct Mail to recruit one-off donors [aka Cash Donors]) has been becoming more difficult. And by more difficult I mean the pay back period has shifted from 12-18 months, out to over 3-4 years, or even more in some cases.
One of the areas that is quite poorly measured and assessed across the sector (or at least to my knowledge) is that of online, particularly in the social space. So armed with my tools I’ve decided to take a look at three charities in Australia to see what they’re doing on Facebook.
The charities I’ve chose have scale in income, are well established and support beneficiaries in the Environment, Health Care & Social Services. What is seen here may not be reflective of ALL charities using Facebook, and given time I might extend the analysis further to incorporate a broader list of charities. I have anonymised them for the purposes of this analysis.
The first thing I wanted to do is to have a look at how often these charities were posting to Facebook. Is it once or twice a day or is it once a week, every other day, etc. I also wanted to get a feel for how often they were asking their fans for a donation. After all whenever we send a letter we typically include an ask of some form. Is it going to be the same for people who’re engaging with us on Social Media?
The chart below shows the number of days between all Facebook posts, and the number of days between posts that mention ‘donate’ or some derivative of the word (yes there is a small chance that we get a false positive E.g. Thanks to all who’ve donated…).
We can clearly see that our three charities are posting at least once a day, if not closer to twice. However when we look at how often charities are asking or even just mentioning donations, the frequency drops to once a week, or in the worst case, less than once a fortnight. This is not to say that they aren’t doing complementary advertising to drive donations, they just don’t appear to be promoting their need for donations to their fan base.
Given what we’ve seen in the chart above, lets simplify that a little bit. So what proportion of posts use the word ‘donate’ or similar (E.g. Donation, donated, donates, etc)?
It is clear that it varies quite widely. Anywhere from 3.2% of posts to 14.1% of posts mention the word ‘donate’, so we might assume have some form of ask. Based on nothing other than a gut feel, 14-20% seems like it might be a good balance of how often we should ask, given the type of audience and the diversity of content that is often served in Facebook pages.
Finally on the frequency of the donation it’s useful to see when posts are made to Facebook, and how that might affect engagement and the like. The chart below looks at the time of day and the date of the post, plotting all the posts and whether or not they contain a ‘donation’ ask.
It becomes pretty clear that the charities really only post during business hours, the environment charity appears to have some scheduled posts for the same time each day and after October 2016 they start to deviate slightly from just before 8am, 12pm and 3pm to have a slightly more organic approach to posting. The Health and Social charities do appear to be much more organic in their approach, however it does appear rare for any of them to be posting before 8am, or beyond 6pm at night. The positive news in this is that the donation posts do appear to be spread across an assortment of times.
The final piece of analysis I wanted to explore was the sentiment of the content being posted. There is often are cases for and against the use of particular language to build empathy or guilt. At this point I don’t have an opinion on whether one is better or worse at encouraging people to give, and what motivations it facilitates in those who give.
I’m more curious to see where on the spectrum of positive or negative does our social media content sit. Further to this, does a particular sentiment create stronger engagement with a cause or not? Firstly lets start with the sentiment. Below is a set of histograms that show where on the sentiment spectrum a post sits. Below 0 is negative, 0 is neutral, greater than 0 is positive.
Most of the posts have a fairly neutral sentiment. Very few are particularly negative, but quite a number have positive tones, or are quite positive.
Now lets have a look at the distribution of likes. The chart below shows a set of histograms indicating the number of likes each post got.
Unfortunately this isn’t an entirely fair measure, as it is evident one of our three charities gets far more likes than the other two combined. Most posts only get a handful of likes, and to more fairly show the distribution across the spectrum (up to 25K likes) we’ve used a
log10 scale on the
Finally I thought it would be worth plotting the above information against our timescales from earlier. The chart below plots the charities Facebook posts over time and colours the dots based on sentiment, and sizes them based on fan engagement (the count of likes in this case).
Unfortunately with the small variance in sentiment for most of the posts it does make it hard to see much colour variation. With regards to our engagement (count of likes) we see the strength of our environment charity in garnering at the very least click support.
The next logical extension of this analysis would be to see whether or not it translates into donations. This may be slightly trickier to do, as I am sure I’d need some willing charities to share their private Facebook data, and possibly their Google Analytics data.
Let me know in the comments if this is something you’d be open to share.